Anthropic Pays $460K for a PM. The Named Pool Is 21 People.

Amol Avasare, Anthropic's Head of Growth, said the quiet part out loud on Lenny's Podcast: since adopting Claude Code, Anthropic's engineering teams are shipping at two to three times their effective headcount, and the growth team's next hires won't be engineers. They'll be product managers. That single sentence flips a 15-year assumption about how you staff a software company, and it lands at the exact moment every AI-native org is about to have the same problem.

The catch: the pool of PMs who can actually do the job barely exists. In Refolk's index of professional profiles, only 21 people in the US and UK hold the literal title "AI Product Manager." Anthropic is paying $305K to $460K for one.

Why Claude Code broke the engineer-to-PM ratio

The traditional 1:8 PM-to-engineer ratio now plays out closer to an effective 1:20 at AI-native orgs, because each engineer ships roughly 8x more code per quarter. The constraint moved from the IDE to the roadmap.

Here is the mechanism. More than 80% of the code merged into Anthropic's production codebase in May was Claude-authored, not human-authored. A five-person engineering team now ships like a team of 15 to 20. When you compress engineering that hard, the bottleneck doesn't disappear, it migrates. It moves to whoever decides what to build, which is a PM function. Avasare's public argument is that the ratio may need to flip entirely: more PMs than engineers, at least in product-led surface areas.

8x

If you are running a 40-person AI-native company today with four PMs and 32 engineers, the honest read is that you are under-PM'd by roughly a factor of 2.5. Not because your PMs are bad, but because your engineers stopped being the constraint sometime in the last 18 months and nobody re-plotted the org chart.

The AI-native product manager pool is smaller than a single PM org

Refolk's index shows 1,022 US and UK profiles with any "AI Product Manager"-style title across all seniority levels, and just 112 US Senior-and-above PMs with "AI" or "LLM" in their headline. The entire named category is smaller than the PM org at a single mid-size public company.

The numbers get worse the more specific you make the query. Search Refolk's index for senior US PMs with "Claude," "LLM," or "foundation model" in their headline and the count drops to zero. The exact skillset Anthropic is buying for a ~$382K midpoint does not yet exist as a labeled category on the open market.

Two numbers do most of the work here. Named "AI Product Manager" titleholders make up roughly 2% of the senior "AI/LLM" PM pool (21 of 1,022). And Anthropic's PM band runs about 1.6x standard Big Tech Senior PM comp. That is not a market rate. That is a scarcity premium in a market where four frontier labs (Anthropic, OpenAI, Google DeepMind, xAI) are fishing the same hundred-person pond.

The frontier-lab AI PM pool hides behind generic titles

The people you actually want don't call themselves AI PMs. They are Senior and Staff PMs at Meta, Airbnb, Walmart, CVS Health, Fiserv, and Equifax who quietly shipped LLM features under generic product titles, and they will not surface in any title-based Boolean search.

In Refolk's index, the top employers of Senior+ PMs with "AI/LLM" headline signal cluster at Meta (three profiles), then Airbnb, Walmart, CVS Health, Fiserv, and Equifax. Notice what's missing: the frontier labs themselves. Anthropic, OpenAI, and DeepMind AI PMs are essentially invisible in the open market because they don't self-tag, and because they don't need to job-hunt. This is why keyword searches for "AI PM" return a wall of tourists and almost none of the real operators.

The sourcing shift is evidence-of-work over titles:

• Public RFCs and design docs on GitHub, Notion, or company engineering blogs.
• Launch posts and postmortems where a PM is named as the DRI on an LLM feature.
• Conference talks (Lenny's Podcast, AI Engineer Summit) with a named PM guest.
• Internal transfers - a PM who moved from Payments to "AI Foundations" 14 months ago is a stronger signal than any headline keyword.
• GitHub activity on prompt libraries, evals repos, or LLM tooling under a real name.

This is the exact gap Refolk closes. You describe the person in plain English ("Senior PM who shipped an LLM feature at a Fortune 500, has a GitHub with eval work, not currently at a frontier lab") and get a ranked shortlist that ignores the title field entirely. Titles are the last thing you should filter on when the category is 18 months old.

What Anthropic's $305K to $460K PM band actually signals

That salary band is not compensation philosophy, it is a supply signal. When a company pays 1.6x the standard Big Tech Senior PM comp band for a PM role, it is telling you the effective supply is near zero and the buyer set has four names on it.

Here is the arithmetic founders need to run. Anthropic raised a $13B Series F led by ICONIQ with Fidelity and Lightspeed co-leading at a $183B valuation, then closed a $30B Series G on February 12, 2026 at a $380B post-money. It plans to triple its global workforce, open its first Bengaluru office in 2026, and grow its applied AI team fivefold by year end. If the effective PM ratio needs to move from 1:8 toward 1:20-with-more-PMs, the four frontier labs alone will absorb a large fraction of the named 112-person senior AI/LLM PM pool in the next 12 months. Every non-frontier AI company will be hiring from what's left, which is mostly the hidden pool at Meta, Airbnb, and Fortune 500 AI teams.

When a category's salary band runs 1.6x the market and its named pool is 21 people, you are not hiring, you are competing in an auction.

The tactical read for founders and heads of talent:

1. Stop budgeting AI PM comp against your existing PM band. Budget against Anthropic's band, minus a founder-story discount if you have one.
2. Assume every strong candidate has two other offers within 30 days of first contact. Move accordingly.
3. Source from adjacent categories (staff engineers who did PM-adjacent work, TPMs on ML platforms, former founders who shipped AI products) before you burn out the named 112.
4. Build an internal path: promote a Senior PM with LLM curiosity now. In 12 months they will be worth $400K on the open market and you will have paid a fraction of that to grow them.

The applied AI hire nobody has a filter for

Anthropic plans to grow its applied AI team fivefold by year end to help customers set up large-scale Claude deployments, and this is the sleeper category: a PM-engineer-solutions hybrid that no job board has a filter for. If you are sourcing for AI teams in 2026, this is the role you will underestimate.

Applied AI engineers sit between the model, the customer, and the roadmap. They write code, they run evals, they scope customer deployments, and they feed insights back to the PM and research teams. They are part of why enterprise revenue is driving roughly 80% of Anthropic's income across 300,000 business and enterprise customers, with Claude Code alone nearing a $1B run rate. Chris Ciauri, newly appointed EVP for International Operations and previously CEO of Unily with earlier stints at Salesforce and Google, is the enterprise-scaling profile Anthropic is now hiring against internationally.

The sourcing problem: there is no clean title. The applied AI hybrid gets tagged as "ML Engineer," "Solutions Architect," "Forward Deployed Engineer," "AI Solutions Engineer," and half a dozen other labels. Boolean searches miss most of the real pool. Refolk closes this cleanly too: describe the shape of the role ("someone who can write production Python, run LLM evals, and sit in a customer's Slack") and the ranked shortlist ignores whichever of the six titles the person happens to carry.

How to rebuild your sourcing pipeline for the 1:20 world

Rebuild around three things: evidence-of-work signals, adjacent-category expansion, and a comp band anchored to Anthropic's, not to your 2024 PM ladder. That is the whole playbook, and most talent teams are still executing the 2023 version of it.

Concretely, here is what changes:

• Ratio target. If you are AI-native, plan for 20 to 30% of headcount in product roles, not 10%. Budget accordingly.
• Title filters. Turn them off. In a category with 21 named titleholders and 1,022 loose-fit profiles, title filters remove your best candidates.
• Comp bands. Anchor Senior AI PM at $280K to $380K total comp minimum, Staff at $350K to $460K. Below that you are noise to the candidate.
• Sourcing channels. Lenny Rachitsky's community (where both Avasare and Cat Wu, Head of Product for Claude Code, have appeared), the Claude Code user community, and the Meta AI PM alumni network - not LinkedIn job posts.
• Non-traditional paths. Avasare cold-emailed Mike Krieger (Anthropic CPO, Instagram co-founder) when no job listing existed, after growth stints at Mercury and MasterClass. Cat Wu was an engineer for years, then briefly in VC, before running product for Claude Code. Neither would pass a keyword filter.
• Evidence-of-work capture. For every candidate, log two public artifacts (RFC, launch post, GitHub repo, podcast, conference talk) before the first call.

The founders who get this right in the next 12 months will build teams that ship faster than their competitors' engineering headcount would suggest is possible. The ones who don't will hire the wrong 32 engineers and wonder why the roadmap keeps slipping.

FAQ

How many AI-fluent product managers actually exist in the US and UK?

In Refolk's index, 1,022 people in the US and UK hold an "AI Product Manager"-style title across all seniority levels, and only 21 hold the literal title "AI Product Manager." The senior-and-above pool with "AI/LLM" in the headline is 112 people in the US. Zero US Senior PMs currently have "Claude" or "foundation model" in their headline, which means the exact skill Anthropic is paying $305K to $460K for is not yet a labeled category on the open market. The real pool is larger, but it hides behind generic PM titles at Meta, Airbnb, and Fortune 500 AI teams.

What is the new engineer-to-PM ratio at AI-native companies?

The historical ratio was roughly 1:8. Since Claude Code drove an 8x increase in code shipped per engineer per quarter at Anthropic, with over 80% of merged code being Claude-authored in May, the effective ratio now sits closer to 1:20. Amol Avasare has publicly argued it may need to flip entirely - more PMs than engineers. For most AI-native orgs, that means planning for 20 to 30% of headcount in product roles rather than the traditional 10%.

Why is Anthropic paying $305K to $460K for a Claude Code PM?

That band runs about 1.6x standard Big Tech Senior PM total comp, which is a scarcity premium, not a market rate. When the named pool is 21 people, the loose pool is 112, and four frontier labs (Anthropic, OpenAI, Google DeepMind, xAI) are all buying at once, prices detach from ladder logic. Founders budgeting AI PM comp against their existing bands will lose every competitive process.

Where should I actually source AI PMs if title searches don't work?

Source on evidence of work, not titles. Look for public RFCs, launch posts where a PM is named as DRI on an LLM feature, GitHub activity on evals or prompt libraries, conference talks (Lenny's Podcast, AI Engineer Summit), and internal transfers into AI teams at Meta, Airbnb, or Fortune 500 companies in the last 18 months. Refolk lets you describe the profile in plain English across LinkedIn, GitHub, and the open web, which is the only way to surface the majority of the real pool that hides behind generic Senior PM titles.